This invention relates generally to delta-sigma analog-to-digital converters. More specifically, the present invention provides circuits and methods for a delta-sigma analog-to-digital converter having a variable oversample ratio produce a constant fullscale output with reduced circuit complexity, die area, and power dissipation.
Analog-to-digital converters (xe2x80x9cADCxe2x80x9d) are electronic devices that convert analog signals into digital representations. As such, they form an integral part of any digital system requiring an interface between external analog signals and the digital circuits in the system.
A block diagram of an ADC is shown in FIG. 1. ADC 20 uses reference voltage Vref to convert analog signal Vin into N-bit digital signal Dout. Analog signal Vin is first sampled into a discrete-time signal and then the discrete-time signal is quantized into a finite number of quantization levels to produce Dout. For an N-bit Dout, Vin is quantized into 2N levels, with each level separated by a quantization step size. As a result of the quantization, a number of input voltage signal levels produces identical digital outputs.
Reference voltage Vref provides the range of conversion for the ADC so that input signal Vin may range from 0 to +Vref or from xe2x88x92Vref to +Vref (for a bipolar ADC) If Vin is equal to or larger than Vref, commonly referred to as the fullscale input, Dout outputs all ones and is referred to as the fullscale output. If Vin is equal to or smaller than 0/xe2x88x92Vref V, Dout outputs all zeros. For Vin between these two voltage levels, Dout is a binary number corresponding to the Vin signal level such that a change in Vin of a quantization step size of Vref/2N corresponds to a 1-bit change in the least significant bit (xe2x80x9cLSBxe2x80x9d) of
The performance of an ADC is evaluated based on its resolution, accuracy, and speed. The resolution of an ADC is determined by the number of bits used to represent Dout. An N-bit ADC has a resolution of 1:2N. The accuracy of the conversion is represented in terms of the quantization step size/bit or in terms of the RMS noise generated for a fixed input. The speed or conversion rate of the ADC is the time it takes for the ADC to perform a conversion. The higher the number of times an input is sampled per conversion result, the higher the resolution and accuracy of the conversion and the slower the speed of the ADC. For example, an 8-bit ADC having a Vref of 5 V quantizes the input voltage into 256 levels with a quantization step size of 19.5 mV. That is, the ADC cannot resolve input voltage differences smaller than 19.5 mV, i.e., this 8-bit ADC has an accuracy of 19.5 mV/bit. In contrast, a 12-bit ADC with 4096 quantization levels can resolve voltage differences as small as 1.2 mV, i.e., its accuracy is 1.2 mV/bit.
The trade-off between resolution, accuracy, and speed of an ADC is highly dependent on its architecture. There are many different architectures of ADCs available, with the most popular ones being the parallel or flash converter, the successive approximation ADC, the voltage-to-frequency ADC, the integrating ADC, and the delta-sigma or sigma-delta ADC. The parallel converter is the simplest and fastest ADC, with the N output bits determined in parallel by 2Nxe2x88x921 comparators. However, because this architecture requires a large number of comparators, commercial parallel ADCs have very limited resolution, up to 1:1024 (10-bit outputs). Examples of commercially available parallel ADCs include the 8-bit ADC0820, sold by National Semiconductor, of Santa Clara, Calif., and the 8-bit AD7820, sold by Analog Devices, Inc., of Northwood, Mass.
Successive approximation ADCs are also relatively fast, employing a digital-to-analog converter (xe2x80x9cDACxe2x80x9d) to try out various digital output levels and a single comparator to compare the result of the DAC conversion to the analog input voltage. For a N-bit successive approximation ADC, N comparisons are required. Successive approximation ADCs are inexpensive to implement and commercial implementations typically range from 8 to 16 bits. Examples of commercially available successive approximation ADCs include the 12-bit LTC1410, sold by Linear Technology Corp., of Milpitas, Calif., and the 8-bit ADC0801, sold by National Semiconductor, of Santa Clara, Calif.
If speed is not important, voltage-to-frequency ADCs offer an inexpensive architecture suitable for converting slow and often noisy signals. These ADCs convert an input voltage into an output pulse train whose frequency is proportional to the input voltage. The output frequency is determined by counting pulses over a fixed time interval. Commercially available voltage-to-frequency ADCs have outputs ranging from 8 to 12 bits and are useful for applications in noisy environments when an output frequency is desired, such as in remote sensing applications when an analog input voltage is converted to an output pulse train at a remote location and the output pulse train is transmitted over a long distance to eliminate the noise introduced in the transmission of an analog signal. Examples of voltage-to-frequency ADCs include the AD650, sold by Analog Devices, Inc., of Northwood, Mass., and the LM331, sold by National Semiconductor, of Santa Clara, Calif.
For low speed applications requiring higher resolution, integrating ADCs provide a better alternative to voltage-to-frequency ADCs. Integrating ADCs measure the charge and discharge times of a capacitor to determine the digital output according to the relationship between the input voltage and the capacitor charge and discharge times. In single-slope integrating ADCs, the relationship is determined by counting clock pulses until a comparator finds the capacitor charged to the input voltage. The digital output is given by the number of clock pulses. In dual-slope integrating ADCs, the relationship is determined by charging the capacitor for a fixed time period with a current that is proportional to the input voltage and subsequently discharging the capacitor with a constant current. The time to discharge the capacitor is proportional to the input voltage and the digital output is given by the number of clock pulses counted while the capacitor is discharging. Single-slope integrating ADCs are simple to implement but not as accurate as dual-slope integrating ADCs, which are commonly used in high precision digital systems. The resolution of commercially available integrating ADCs may range from 1:210 to 1:220. Examples include the 18-bit ALD500, sold by Advanced Linear Devices, Inc., of Sunnyvale, Calif., and the 18-bit AD1170, sold by Analog Devices, Inc., of Northwood, Mass.
Although the ADC architectures discussed above provide a wide range of choices in terms of resolution, accuracy, and speed, their analog components make it difficult to integrate their circuitry in high-speed VLSI technology. Because they operate at a relatively low sampling frequency, usually at the Nyquist rate of the input signal, they often require an external anti-aliasing analog filter and sample-and-hold circuitry to limit the frequency of the input signal. Additionally, these ADC architectures are vulnerable to noise and interference and require high-accuracy analog components in order to achieve high resolution.
Currently available delta-sigma ADCs provide a solution to the VLSI integration and noise problems of the previous ADC architectures. Delta-sigma ADCs use a low resolution (e.g., 1-bit) delta-sigma analog modulator running at very high sampling rates combined with a digital filter to achieve high output resolutions. The modulator oversamples the input signal, transforming it into a serial bit stream at a frequency well above the output rate. The digital filter then low-pass filters and decimates the bit stream generated by the modulator to achieve an improved resolution at a lower output rate. For example, a 20-bit delta-sigma ADC may be implemented by combining a 1-bit delta-sigma modulator sampling an input multiple times and applying the result to a digital filter. Since a 1-bit delta-sigma modulator does not require special analog circuit processes, the delta-sigma ADCs can be easily implemented into VLSI technology and integrated into complex monolithic systems that incorporate both analog and digital components. The implementation cost is low and will continue to decrease with further advances in VLSI technology.
Additionally, as a result of the higher input sampling rate, delta-sigma ADCs require a much simpler anti-aliasing analog filter than traditional ADCs and no external sample-and-hold circuitry. The digital filter can be tailored to minimize the noise as desired. Commercially available delta-sigma ADCs also achieve higher resolutions than the other ADC architectures discussed above, with the resolutions typically ranging from 1:216 to 1:224. Delta-sigma ADCs are increasingly replacing voltage-to-frequency and integrating ADCs as the preferred architecture in many applications. Examples of delta-sigma ADCs include the 24-bit LTC2400 and the 24-bit LTC2410, sold by Linear Technology Corp., of Milpitas, Calif.
A block diagram of a delta-sigma ADC is shown in FIG. 2. Delta-sigma ADC 25 consists of two components: oversampled analog delta-sigma modulator 30 and low-pass digital filter 35. Oversampled analog modulator 30 samples the input signal at a sampling rate Fsample that is much higher than the Nyquist frequency to produce a B-bit stream of data. As a result, the quantization noise is high-pass noise shaped over a bandwidth equal to Fsample so that most of the energy of the quantization noise is above the bandwidth of the input signal. The quantization noise is then filtered out by low-pass digital filter 35, which also performs a decimation step to produce a M-bit digital output, with M greater than  greater than B, at a sampling rate of Fout less than  less than Fsample. The ratio between oversampled analog modulator 30 sampling rate Fsample and the sampling rate Fout, which is ADC 25""s conversion rate, is referred to as the oversample ratio (xe2x80x9cOSRxe2x80x9d), that is, OSR=Fsample/Fout.
The oversample ratio represents the number of times the input signal is sampled for each analog-to-digital conversion. As OSR increases, the number of times the input signal is sampled increases, thereby decreasing the passband noise output by modulator 30. The reduction in noise at the lower frequencies combined with digital filter 35 increases the resolution of ADC 25. Increasing OSR for a given sampling rate Fsample also decreases Fout, i.e., the speed of ADC 25 decreases. That is, OSR offers a trade-off between speed and resolution.
To achieve the different resolution and accuracy requirements of a wide range of applications, multiple ADCs running at different speeds and resolutions may be used. Preferably, a single ADC may be used if it is designed to handle a variable OSR. Having a variable OSR in a delta-sigma ADC implies that the size of the digital filter of a given filter order is determined by the maximum allowed OSR. Such an ADC can run at different resolutions and speeds but requires further digital processing to produce conversion results that are independent of OSR. For example, an M-bit digital filter will only be able to produce a fullscale digital output for a fullscale input when the OSR is at its maximum. If the OSR is reduced such that the ADC""s resolution is reduced from 1:2M to 1:2j, where J less than M, the M-bit output will only have J ones and the top Mxe2x88x92J bits will be equal to zero and unused. This implies that the conventional delta-sigma ADC architecture shown in FIG. 2 is not ideally suited to handle a variable OSR as it wastes die area, circuitry complexity, and power to generate unused bits when the OSR is lower than the maximum allowed OSR.
In view of the foregoing, it would be desirable to provide circuits and methods for a delta-sigma analog-to-digital converter to handle a variable oversample ratio that provides various resolutions and conversion rates.
It further would be desirable to provide circuits and methods for a delta-sigma analog-to-digital converter to offer various resolutions and conversion rates with reduced design complexity, die area, and power dissipation.
It also would be desirable to provide circuits and methods for a delta-sigma analog-to-digital converter to produce a constant fullscale output independent of its oversample ratio.
In view of the foregoing, it is an object of the present invention to provide circuits and methods for a delta-sigma analog-to-digital converter to handle a variable oversample ratio that provides various resolutions and conversion rates.
It is a further object of the present invention to provide circuits and methods for a delta-sigma analog-to-digital converter to offer various resolutions and conversion rates with reduced design complexity, die area, and power dissipation.
It is also an object of the present invention to provide circuits and methods for a delta-sigma analog-to-digital converter to produce a constant fullscale output independent of its oversample ratio.
These and other objects of the present invention are accomplished by providing circuits and methods for a delta-sigma analog-to-digital converter having a variable oversample ratio to produce a constant fullscale output at reduced circuit complexity, die area, and power dissipation.
The circuits and methods of the present invention consist of implementing the digital filter in the delta-sigma converter as a comb filter and scaling the input to the comb filter. In a preferred embodiment, a decoder is used-to scale the input to the comb filter. The decoder adjusts the B-bit output provided by the oversampled analog modulator according to its OSR so that the fullscale output of the digital filter is independent of the OSR. Consequently, there are no wasted bits and the lower bits output by the digital filter are always zero. Since the filter size is much larger than the required resolution, the lower bits may be removed, that is, the hardware required to output the lower bits need not be implemented.
The comb filter is implemented in hardware as a cascade of integrators and differentiators and the decoder is implemented as a 1:J decoder, where J is the number of OSRs allowed by the delta-sigma converter.
Advantageously, the present invention enables the delta-sigma converter to produce a constant fullscale output independent of the OSR. The present invention also reduces circuitry complexity, die area, and power dissipation of a delta-sigma converter when compared to other previously-known delta-sigma converters that produce a constant fullscale output independent of the OSR.